Image

Visualizing U.S. Commuters

I recently analyzed the patterns of U.S. Commuters and created a visualization that summarizes these patterns at the state level.

About the Data

This data is from the Census Transporation Planning Products (CTPP). For those who don’t know, the CTPP is derived from the 2006–2010 5-year American Community Survey (ACS) data. You can learn more about this data set by visiting the home page. In an effort to make this easily reproducible, you can download the csv used in this analysis. I chose this data source because I have used it in the past and am familiar with it.

Steps Taken

I created a directed network graph from this data. I used python’s networkx and pandas packages and the complete source code is provided below. I excluded Puerto Rico from the data set because I wanted to analyze state level patterns. I did leave the District of Columbia in thus there are 51 States in the analysis.

I wanted to use Gephi to generate the visualization because I have done so in the past. After creating the directed network graph using Python, I imported the data into Gephi and used the community detection algorithm using the default settings (Randomized checked, Use weights checked, Resolution = 1.0). This resulted in 7 communities being detected.

I grouped the states by these communities and colored them and placed them in a circular layout with straight edges between the nodes. I varied the width based on the edge weight.

Visualization

Commuters Network

Observations

There are a couple of things that are of interest. The first thing to acknowledge is the proliferation of edges in this network.  Almost all of the states are connected with the other states.  This results in the “spirograph” like effect in the visualization.  However I don’t find that to be the most interesting aspect.

The sub-networks highlighted in this visualization are particularly interesting.  For example one readily sees that a lot of people in New Jersey work in New York.  As a former New Yorker there is no surprises there.  You also see the Capital Beltway connections in the visualization.  Residents of Maryland and Virginia find work in D.C.  The community detection algorithm highlights this finding.  Since the states are arranged by “community”, seeing the cross-community connections are interesting.  For example the Connecticut to New York have a connection.

Source Code

The following is how I generated the directed network graph file for Gephi from the CSV.

import pandas as pd
import networkx as nx

df = pd.read_csv('Table 1 Commuting Flows Available at State to POW State and County to POW County only.csv', skiprows=[0,1], thousands=',')
# Only pull the Estimates
df = df[df['Output']=='Estimate']
# Only pull the first 4 columns
df = df[df.columns[:4]]
# Drop the N/A's
df = df.dropna()
# Drop the Output column
df = df.drop('Output', 1)
# Rename the columns
df.columns = ['from','to','weight']
# Drop the Puerto Rico records
df = df[df['from'] != 'Puerto Rico']
df = df[df['to'] != 'Puerto Rico']
# Remove the people who work where they live
commuters = df[df['from'] != df['to']]
# Build the network graph
G = nx.from_pandas_dataframe(commuters, 'from','to', ['weight'], create_using=nx.DiGraph())
# Write the graph so I can use Gephi
nx.write_gexf(G,'Commuters.gexf')

K Means Clustering (Week 4 Assignment)

Introduction

A k-means clustering analysis was conducted to divide countries into subgroups based on variables that were determined to have an impact on economic well-being. The data for the analysis is and extract from the GapMinder project. The GapMinder project collects country-level time series data on health, wealth and development. The data set for this analysis only has one year of data for 213 countries.

Potential Clustering Variables

A random forest analysis was preformed on a training set (N=111) to evaluate a series of explanatory variables in predicting a categorical GDP per person binned into thirds. The following explanatory variables were evaluated:

  • Alcohol Consumption – 2008 recorded and estimated average alcohol consumption, adult (15+) per capita as collected by the World Heath Organization
  • CO2 Emissions – Total amount of CO2 emission in metric tons from 1751 to 2006 as collected by CDIAC
  • Female Employment Rate – Percentage of female population, age above 15, that has been
    employed during 2007 as collected by the International Labour Organization
  • Internet Use Rate – 2010 Internet users per 100 people as collected by the World Bank
  • Life Expectancy – 2011 life expectancy at birth (in years) as collected by various sources
  • Polity Score – 2009 Democracy score as collected by the Polity IV Project
  • Employment Rate – Percentage of total population, age above 15, that has been employed during 2009 as collected by the International Labour Organization
  • Urbanization Rate – 2008 Urban population (% total population) as collected by the World Bank

The following table summarizes the variables importance in classifying countries into high middle and low GDP per person countries:

Table 1. Variables Importance

Variable Importance
Life Expectancy 35%
Internet Use Rate 21%
Urbanization Rate 11%
Female Employ 8%
CO2 Emissions 8%
Employ Rate 6%
Alcohol Consumption 6%
Polity Score 5%

The explanatory variables with a relative importance scores greater than or equal to 10% (life expectancy, internet use rate and urbanization rate) we selected as clustering variables.

K-Means Clustering

All clustering variables were standardized to have a mean of 0 and a standard deviation of 1. Data were randomly split into a training set that included 70% of the observations (N=121) and a test set that included 30% of the observations (N=52). A series of k-means cluster analyses were conducted on the training data specifying k=1-9 clusters, using Euclidean distance. The mean Euclidean distance was plotted for each of the nine cluster solutions in an elbow curve to provide guidance for selecting the number of clusters. The following figure illustrates the elbow curve

Figure 1. Selecting K with the Elbow Curve Method

elbow

Possible levels of k include 2, 3, 4 and 7. Each level was evaluated to see if there was a significant difference between levels of economic well-being (GDP per person). Breaking the data into 2 clusters proved to be a good choice however I hasten to note that 7 was not a bad choice. My preference towards less complication than was needed lead to selecting k=2.

Findings

Cluster number zero has a much lower economic well-being per person than cluster one.

Table 2. GDP Per Person by Cluster

Cluster Average GDP per Person
Zero $1,373
One $15,684

The following figure shows the clusters (zero in blue, one in green) and the clustering variables:

Figure 2. Clustering Variables Scaterplots

k-means pairplot

Discussion

This examination was done as an exercise in doing K-means clustering. The data in the GapMinder dataset does not lend itself to clustering. The scatter plots above illustrate the lack of distinct clusters. Caution should be exercised in using/interpreting these results.

Source Code

As always this project can be found in this GitHub repository.

# Import libraries needed
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from sklearn.cross_validation import train_test_split
from sklearn import preprocessing
from sklearn.cluster import KMeans
from scipy.spatial.distance import cdist
import statsmodels.formula.api as smf
import statsmodels.stats.multicomp as multi 
import seaborn as sns
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier
sns.set_style('whitegrid')
sns.set_context('talk')

# Eliminate false positive SettingWithCopyWarning
pd.options.mode.chained_assignment = None

# Make results reproducible
np.random.seed(1234567890)

df = pd.read_csv('gapminder.csv')

variables = ['incomeperperson', 'alcconsumption', 'co2emissions', 'femaleemployrate', 
                'internetuserate', 'lifeexpectancy','polityscore','employrate','urbanrate'] 

# convert to numeric format
for variable in variables:
    df[variable] = pd.to_numeric(df[variable], errors='coerce')

# listwise deletion of missing values
subset = df[variables].dropna()

# Print the rows and columns of the data frame
print('Size of study data')
print(subset.shape)
print("\n")
"""
"===============  Random Forest to Select Clustering Variables  ===============
"""
n_estimators=25

subset['incomequartiles'] = pd.cut(subset['incomeperperson'], 3, labels=['0%-33%','34%-66%','67%-100%'])
subset['incomequartiles'] = subset['incomequartiles'].astype('category')

variables.pop(0)

predictors = subset[variables]
targets = subset['incomequartiles']

#Split into training and testing sets+
training_data, test_data, training_target, test_target  = train_test_split(predictors, targets, test_size=.25)

# Build the random forest classifier
classifier=RandomForestClassifier(n_estimators=n_estimators)
classifier=classifier.fit(training_data,training_target)

predictions=classifier.predict(test_data)

# Fit an Extra Trees model to the data
model = ExtraTreesClassifier()
model.fit(training_data,training_target)

# Display the relative importance of each attribute
feature_name = list(predictors.columns.values)
feature_importance = list(model.feature_importances_)
features = pd.DataFrame({'name':feature_name, 'importance':feature_importance}).sort_values(by='importance', ascending=False)
print(features.head(len(feature_name)))

"""
" =============================  Data Management  =============================
"""
variables = ['incomeperperson', 'lifeexpectancy', 'internetuserate', 'urbanrate'] 

# convert to numeric format
for variable in variables:
    df[variable] = pd.to_numeric(df[variable], errors='coerce')

# listwise deletion of missing values
subset = df[variables].dropna()

# Print the rows and columns of the data frame
print('Size of study data')
print(subset.shape)

subset['incomequartiles'] = pd.cut(subset['incomeperperson'], 3, labels=['0%-33%','34%-66%','67%-100%'])
subset['incomequartiles'] = subset['incomequartiles'].astype('category')

# Remove the first variable from the list since the target is derived from it
variables.pop(0)

# Center and scale data
for variable in variables:
    subset[variable]=preprocessing.scale(subset[variable].astype('float64'))
    
features = subset[variables]
targets = subset[['incomeperperson']]

"""
" ==================  Split Data into Training and Test Sets  ==================
"""
# Split into training and testing sets
training_data, test_data, training_target, test_target  = train_test_split(features, targets, test_size=.3)
print('Size of training data')
print(training_data.shape)

"""
" =====================  Determine the Number of Clusters  ====================
"""
# Identify number of clusters using the elbow method
clusters=range(1,10)
meandist=[]

for k in clusters:
    model=KMeans(n_clusters=k)
    model.fit(training_data)
    clusassign=model.predict(training_data)
    meandist.append(sum(np.min(cdist(training_data, model.cluster_centers_, 'euclidean'), axis=1)) / training_data.shape[0])

# Visualize the elbow
k = 2

fig = plt.figure()
ax = fig.add_subplot(111)
plt.plot(clusters, meandist)
ax.plot(clusters[(k-1)], meandist[(k-1)], marker='o', markersize=12, 
    markeredgewidth=2, markeredgecolor='r', markerfacecolor='None')
plt.grid(True)
plt.xlabel('Number of Clusters')
plt.ylabel('Average Distance')
plt.title('Selecting K with the Elbow Method')
plt.show()

"""
" ==========================  Visualize the Clusters  =========================
"""
model=KMeans(n_clusters=k)
model.fit(training_data)
training_data['cluster'] = model.labels_

my_cmap = plt.cm.get_cmap('brg')
my_cmap.set_under('w')

fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(training_data.iloc[:,0], training_data.iloc[:,1], training_data.iloc[:,2], c=training_data['cluster'], cmap=my_cmap)
ax.set_xlabel(training_data.columns.values[0])
ax.set_ylabel(training_data.columns.values[1])
ax.set_zlabel(training_data.columns.values[2])
plt.show()
 
sns.pairplot(training_data, hue ='cluster');

"""
" ====================  Examine Differences Between Clusters  =================
"""

training_target['cluster'] = model.labels_
income_model = smf.ols(formula='incomeperperson ~ C(cluster)', data=training_target).fit()
print (income_model.summary())

print ('means for features by cluster')
m1= training_target.groupby('cluster').mean()
print (m1)

print ('standard deviations for features by cluster')
m2= training_target.groupby('cluster').std()
print (m2)

mc1 = multi.MultiComparison(training_target['incomeperperson'], training_target['cluster'])
res1 = mc1.tukeyhsd()
print(res1.summary())

Lasso Regression (Week 3 Assignment)

Introduction

A lasso regression analysis was conducted to identify a subset of variables from a pool of 8 quantitative predictor variables that best predicted a binary response variable measuring the presence of high per capita income. The data for the analysis is and extract from the GapMinder project. The GapMinder project collects country-level time series data on health, wealth and development. The data set for this analysis only has one year of data for 213 countries.

High per Capita Income (Response Variable)

The 2010 Gross Domestic Product per capita is was classified into high income for cases where the absolute deviation divided by the mean absolute deviation is greater than 3.  The GDP per capita is measured in constant 2000 U.S. dollars and was originally came from the World Bank’s Work Development Indicators.

Explanatory Variables

All explanatory variables were standardized to have a mean of zero and a standard deviation of one. The following explanatory variables were included in the data set:

  • Alcohol Consumption – 2008 recorded and estimated average alcohol consumption, adult (15+) per capita as collected by the World Heath Organization
  • CO2 Emissions – Total amount of CO2 emission in metric tons from 1751 to 2006 as collected by CDIAC
  • Female Employment Rate – Percentage of female population, age above 15, that has been
    employed during 2007 as collected by the International Labour Organization
  • Internet Use Rate – 2010 Internet users per 100 people as collected by the World Bank
  • Life Expectancy – 2011 life expectancy at birth (in years) as collected by various sources
  • Polity Score – 2009 Democracy score as collected by the Polity IV Project
  • Employment Rate – Percentage of total population, age above 15, that has been employed during 2009 as collected by the International Labour Organization
  • Urbanization Rate – 2008 Urban population (% total population) as collected by the World Bank

Methodology

Data were randomly split into a training set that included 70% of the observations (N=148) and a test set that included 30% of the observations (N=45). The least angle regression algorithm with k=10 fold cross validation was used to estimate the lasso regression model in the training set, and the model was validated using the test set. The change in the cross validation average (mean) squared error at each step was used to identify the best subset of predictor variables.

Figure 1. Mean squared error on each fold

lasso

Findings

Of the 8 predictor variables, 3 were retained in the selected model. Internet user rate was the most strongly associated with high income. CO2 emissions and employment rate were also positively associated with high income. These 3 variables accounted for 44.2% of the variance in the high per capita income response variable in the test data set.

Table 1. Regression Coefficients

Variable Regression Coefficients
Internet Use Rate 0.157313
CO2 Emissions 0.019681
Employ Rate 0.001853
Alcohol Consumption 0
Female Employ 0
Life Expectancy 0
Polity Score 0
Urbanization Rate 0

Discussion

The previous week I examined the same variables using a Random Forest. The explanatory variables with the highest relative importance scores were life expectancy, internet use rate, urbanization rate. The LASSO regression only selected the internet use rate as a significant variable. As noted in last week’s post, these findings suggest that my previous work looking at the relationship between the level of democratization and the economic well-being may have been confounded by other variables.

Source Code

# Import libraries needed
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cross_validation import train_test_split
from sklearn.linear_model import LassoLarsCV
from sklearn import preprocessing

# Make results reproducible
np.random.seed(1234567890)

df = pd.read_csv('gapminder.csv')

variables = ['incomeperperson', 'alcconsumption', 'co2emissions', 'femaleemployrate',
                'internetuserate', 'lifeexpectancy','polityscore','employrate','urbanrate']

# convert to numeric format
for variable in variables:
    df[variable] = pd.to_numeric(df[variable], errors='coerce')

# listwise deletion of missing values
subset = df[variables].dropna()

# Print the rows and columns of the data frame
print('Size of study data')
print(subset.shape)
print("\n")
"""
" =============================  Data Management  =============================
"""
# Remove the first variable from the list since the target is derived from it
variables.pop(0)

# Center and scale data
for variable in variables:
    subset[variable]=preprocessing.scale(subset[variable].astype('float64'))
    
# Identify contries with a high level of income using the MAD (mean absolute deviation) method
subset['absolute_deviations'] = np.absolute(subset['incomeperperson'] - np.median(subset['incomeperperson']))
MAD = np.mean(subset['absolute_deviations'])

# This function converts the income per person absolute deviations to a high income flag
def high_income_flag(absolute_deviations):
    threshold = 3
    if (absolute_deviations/MAD) > threshold:
        return 1
    else:
        return 0

subset['High Income'] = subset['absolute_deviations'].apply(high_income_flag)

"""
" ==========================  Build LASSO Regression  ==========================
"""
predictors = subset[variables]
targets = subset['High Income']

#Split into training and testing sets
training_data, test_data, training_target, test_target  = train_test_split(predictors, targets, test_size=.3)

# Build the LASSO regression model
model=LassoLarsCV(cv=10, precompute=False).fit(training_data, training_target)

"""
" ==========================  Evaluate LASSO Model  ============================
"""
# print variable names and regression coefficients
feature_name = list(predictors.columns.values)
feature_coefficient = list(model.coef_)
features = pd.DataFrame({'Variable':feature_name, 'Regression Coefficients':feature_coefficient}).sort_values(by='Regression Coefficients', ascending=False)
print(features.head(len(feature_name)))

#print(dict(zip(predictors.columns, model.coef_)))

# plot coefficient progression
m_log_alphas = -np.log10(model.alphas_)
ax = plt.gca()
plt.plot(m_log_alphas, model.coef_path_.T)
plt.axvline(-np.log10(model.alpha_), linestyle='--', color='k', label='alpha CV')
plt.ylabel('Regression Coefficients')
plt.xlabel('-log(alpha)')
plt.title('Regression Coefficients Progression for Lasso Paths')

# plot mean square error for each fold
m_log_alphascv = -np.log10(model.cv_alphas_)
plt.figure()
plt.plot(m_log_alphascv, model.cv_mse_path_, ':')
plt.plot(m_log_alphascv, model.cv_mse_path_.mean(axis=-1), 'k', label='Average across the folds', linewidth=2)
plt.axvline(-np.log10(model.alpha_), linestyle='--', color='k', label='alpha CV')
plt.legend()
plt.xlabel('-log(alpha)')
plt.ylabel('Mean squared error')
plt.title('Mean squared error on each fold')


# MSE from training and test data
from sklearn.metrics import mean_squared_error
train_error = mean_squared_error(training_target, model.predict(training_data))
test_error = mean_squared_error(test_target, model.predict(test_data))
print ('training data MSE')
print(train_error)
print ('test data MSE')
print(test_error)

# R-square from training and test data
rsquared_train=model.score(training_data, training_target)
rsquared_test=model.score(test_data, test_target)
print ('training data R-square')
print(rsquared_train)
print ('test data R-square')
print(rsquared_test)

Running a Random Forest (Week 2 Assignment)

Introduction

A random forest analysis was preformed to evaluate a series of explanatory variables in predicting a binary categorical variable. The data for the analysis is and extract from the GapMinder project. The GapMinder project collects country-level time series data on health, wealth and development. The data set for this analysis only has one year of data for 213 countries.

High Income (Response Variable)

The 2010 Gross Domestic Product per capita is was classified into high income for cases where the absolute deviation divided by the mean absolute deviation is greater than 3.  The GDP per capita is measured in constant 2000 U.S. dollars and was originally came from the World Bank’s Work Development Indicators.

Explanatory Variables

The following explanatory variables were evaluated:

  • Alcohol Consumption – 2008 recorded and estimated average alcohol consumption, adult (15+) per capita as collected by the World Heath Organization
  • CO2 Emissions – Total amount of CO2 emission in metric tons from 1751 to 2006 as collected by CDIAC
  • Female Employment Rate – Percentage of female population, age above 15, that has been
    employed during 2007 as collected by the International Labour Organization
  • Internet Use Rate – 2010 Internet users per 100 people as collected by the World Bank
  • Life Expectancy – 2011 life expectancy at birth (in years) as collected by various sources
  • Polity Score – 2009 Democracy score as collected by the Polity IV Project
  • Employment Rate – Percentage of total population, age above 15, that has been employed during 2009 as collected by the International Labour Organization
  • Urbanization Rate – 2008 Urban population (% total population) as collected by the World Bank

Findings

The explanatory variables with the highest relative importance scores were life expectancy, internet use rate, urbanization rate.

Table 1 – Variables Importance

Variable Importance
Life Expectancy 40%
Internet Use Rate 21%
Urbanization Rate 10%
CO2 Emissions 7%
Female Employ 7%
Alcohol Consumption 7%
Employ Rate 5%
Polity Score 4%

The accuracy of the random forest was 97%, with the subsequent growing of multiple trees beyond 3, adding little to the overall accuracy of the model.

Discussion

These findings suggest that my previous work looking at the relationship between the level of democratization and the economic well-being may have been confounded by other variables. The level of democratization was assumed to be a cause while some of these explanatory variables (i.e. life expectancy, internet use rate) are more of outcomes that would be correlated to the level of income.

Source Code

As always my project is available on it’s GitHub repository.

# Import libraries needed
import pandas as pd
import numpy as np
import sklearn as sk
import matplotlib.pyplot as plt
from sklearn.cross_validation import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier

# Make results reproducible
np.random.seed(1234567890)

df = pd.read_csv('gapminder.csv')

variables = ['incomeperperson', 'alcconsumption', 'co2emissions', 'femaleemployrate', 
                'internetuserate', 'lifeexpectancy','polityscore','employrate','urbanrate']
                
# convert to numeric format
for variable in variables:
    df[variable] = pd.to_numeric(df[variable], errors='coerce')
    
# listwise deletion of missing values
subset = df[variables].dropna()

# Print the rows and columns of the data frame
print('Size of study data')
print(subset.shape)

"""
" =============================  Data Management  =============================
"""
# Identify contries with a high level of income using the MAD (mean absolute deviation) method
subset['absolute_deviations'] = np.absolute(subset['incomeperperson'] - np.median(subset['incomeperperson']))
MAD = np.mean(subset['absolute_deviations'])

# This function converts the income per person absolute deviations to a high income flag
def high_income_flag(absolute_deviations):
    threshold = 3
    if (absolute_deviations/MAD) > threshold:
        return "Yes"
    else:
        return "No"

subset['High Income'] = subset['absolute_deviations'].apply(high_income_flag)
subset['High Income'] = subset['High Income'].astype('category')

"""
" ===========================  Build Random Forest  ===========================
"""
# Remove the first variable from the list since the target is derived from it
variables.pop(0)

predictors = subset[variables]
targets = subset['High Income']

#Split into training and testing sets+
training_data, test_data, training_target, test_target  = train_test_split(predictors, targets, test_size=.4)

# Build the random forest classifier
classifier=RandomForestClassifier(n_estimators=25)
classifier=classifier.fit(training_data,training_target)

"""
" =========================  Evaluate Random Forest  ==========================
"""
predictions=classifier.predict(test_data)

print('Classification Report')
print(sk.metrics.classification_report(test_target, predictions))

print('Confusion Matrix')
print(sk.metrics.confusion_matrix(test_target, predictions))

print('Accuracy Score')
print(sk.metrics.accuracy_score(test_target, predictions))

# Fit an Extra Trees model to the data
model = ExtraTreesClassifier()
model.fit(training_data,training_target)

# Display the relative importance of each attribute
feature_name = list(predictors.columns.values)
feature_importance = list(model.feature_importances_)
features = pd.DataFrame({'name':feature_name, 'importance':feature_importance}).sort_values(by='importance', ascending=False)
print(features.head(len(feature_name)))

"""
" ========================  Evaluate Number of Trees  =========================
"""
trees = range(n_estimators)
accuracy = np.zeros(n_estimators)

for idx in range(len(trees)):
    classifier=RandomForestClassifier(n_estimators=idx+1)
    classifier=classifier.fit(training_data,training_target)
    predictions=classifier.predict(test_data)
    accuracy[idx] = sk.metrics.accuracy_score(test_target, predictions)
    
plt.cla()
plt.plot(trees, accuracy)

Machine Learning for Data Analysis Week 1

Introduction

Previous research using the GapMinder data set has suggested that a high income per person is a function of the level of democratization and urbanization in a country.  I will preform a decision tree analysis to test the nonlinear relationships between the binary categorical response variable (high income) and these two explanatory variables.

Figure 1 – High Income by Urbanization and Level of Democracy
DecisionTreePlot

About the Data

The data is from the GapMinder project. The GapMinder project collects country-level time series data on health, wealth and development. The data set for this class only has one year of data for 213 countries. There are 155 countries with complete data for the following variables of interest.

Measures

High Income (Response Variable)

The 2010 Gross Domestic Product per capita is was classified into high income for cases where the absolute deviation divided by the mean absolute deviation is greater than 3.  The GDP per capita is measured in constant 2000 U.S. dollars and was originally came from the World Bank’s Work Development Indicators.

Is Full Democracy (Explanatory Variable)

The level of democracy is measured by the 2009 polity score developed by the Polity IV Project. This value ranges from -10 (autocracy) to 10 (full democracy).  The following plot shows the relationship between the response variable and the Polity IV Score.

Figure 2 – Income per Person by Level of Democracy

HighIncomeByGDPandPolity

I collapsed these 21 categories into two categories:

  • Full Democracy (polity score = 10)
  • Not a Full Democracy (polity score =9 to -10)

Thirty-two of the countries are full democracies and the remaining 123 are not.

Urbanization Rate Quartile (Explanatory Variable)

The urbanization rate is measured by the share of the 2008 population living in an urban setting. The urban population is defined by national statistical offices.  This data was originally produced by the World Bank.  This variable was binned into quartiles.  The following plot shows the relationship between the response variable and the level of urbanization:

Figure 3 – Income per Person by Urbanization Rate

HighIncomeByGDPandUrbanRate

Decision Tree Model

The sample of 155 countries were divided into training and test sets using a 60/40 split.  The following image is the decision tree that my model generated on the training set (apologies for the blurriness.  Click on the image for a clearer one).

Figure 4 – High Income per Person Decision Tree Model

DecisionTree

Please keep in mind that the high income response variable is coded “No” or “Yes.”  Consequently the value in the top most box ([77,16]) represents 77 countries with a “No” and 16 with a “Yes.”

The binary “is full democracy” (1 if country is a full democracy or 0 if they are not) was the first variable to separate the sample into two subgroups.  From there the level of urbanization was used to further break down the sample.  Roughly 7% of the countries that are not full democracies (Is Full Democracy <= 0.5 is true) are classified as high income countries.  These high income countries are all in the third and fourth urbanization rate quartiles.

52% of the “full democracy” countries are classified high income countries.  73% of them are in the third and fourth urbanization quartiles and the remaining 27% are in the second quartile.

Model Accuracy

This model was tested on a set of 62 observations.  The total model classified 97% of the test set correctly, with a precision and recall of 75% for classifying as a high income country and 98% for a not a high income country.

Table 1 – Confusion Matrix

Predicted
Actual No Yes
No 57 1
Yes 1 3

Source Code

As always my project is available on it’s GitHub repository.

# Import libraries needed
import pandas as pd
import numpy as np
import sklearn as sk
from sklearn.cross_validation import train_test_split
from sklearn.tree import DecisionTreeClassifier
from IPython.display import Image
from io import BytesIO
import pydotplus as pdp

# Make results reproducible
np.random.seed(0)

# bug fix for display formats to avoid run time errors
pd.set_option('display.float_format', lambda x:'%.2f'%x)

df = pd.read_csv('gapminder.csv')

# convert to numeric format
df['incomeperperson'] = pd.to_numeric(df['incomeperperson'], errors='coerce')
df['polityscore'] = pd.to_numeric(df['polityscore'], errors='coerce')
df['urbanrate'] = pd.to_numeric(df['urbanrate'], errors='coerce')

# listwise deletion of missing values
subset = df[['incomeperperson', 'polityscore', 'urbanrate']].dropna()

# Summarize the data
print(subset[['incomeperperson', 'urbanrate']].describe())

# Identify contries with a high level of income using the MAD (mean absolute deviation) method
subset['absolute_deviations'] = np.absolute(subset['incomeperperson'] - np.median(subset['incomeperperson']))
MAD = np.mean(subset['absolute_deviations'])

# This function converts the income per person absolute deviations to a high income flag
def high_income_flag(absolute_deviations):
    threshold = 3
    if (absolute_deviations/MAD) > threshold:
        return "Yes"
    else:
        return "No"

subset['High Income'] = subset['absolute_deviations'].apply(high_income_flag)
subset['High Income'] = subset['High Income'].astype('category')

# This function converts the polity score to a category
def convert_polityscore_to_category(polityscore):
    if polityscore == 10:
        return 1
    else:
        return 0

# Now we can use the function to create the new variable
subset['Is Full Democracy'] = subset['polityscore'].apply(convert_polityscore_to_category)
subset['Is Full Democracy'] = subset['Is Full Democracy'].astype('category')

# Bin urban rate into quartiles
subset['Urban Rate Quartile'] = pd.qcut(subset['urbanrate'], 4, labels=False)

#Split into training and testing sets
predictors = subset[['Is Full Democracy','Urban Rate Quartile']]
targets = subset[['High Income']]
training_data, test_data, training_target, test_target  = train_test_split(predictors, targets, test_size=.4)

#Build model on training data
classifier=DecisionTreeClassifier()
classifier=classifier.fit(training_data, training_target)

# Check how well the classifier worked
predictions=classifier.predict(test_data)
print(sk.metrics.confusion_matrix(test_target,predictions))

print(sk.metrics.accuracy_score(test_target, predictions))

print(sk.metrics.classification_report(test_target, predictions))

#Displaying the decision tree
out = BytesIO()
sk.tree.export_graphviz(classifier, out_file=out, feature_names=predictors.columns)
graph=pdp.graph_from_dot_data(out.getvalue())
Image(graph.create_png())

Regression Modeling in Practice Week 4 Assignement

Introduction

In this week’s class we will be using logistic regression. I have been examining the relationship between economic well-being and the level of democracy or openness of a society.  My hypothesis is that countries that are the most open will be more likely to have a high level of economic well-being. My previous work has established a statistically significant positive relationship between the level of urbanization and economic well-being. So we will be examining the influence of this confounding variable.

About the Data

The data is from the GapMinder project. The GapMinder project collects country-level time series data on health, wealth and development. The data set for this class only has one year of data for 213 countries. There are 155 countries with complete data for the following variables of interest.

Measures

Economic Well-Being (Response Variable)

Income per person (economic well-being) is the 2010 Gross Domestic Product per capita is measured in constant 2000 U.S. dollars. This allows for comparison across countries with different costs of living. This data originally came from the World Bank’s Work Development Indicators.

Since we are doing a logistic regression I have binned this variable into those that have higher than average income (n=39) and those with less than average incomes (n=116). The average per capita income of the study group is $6,605.

Level of Democracy (Explanatory Variable)

The level of democracy is measured by the 2009 polity score developed by the Polity IV Project. This value ranges from -10 (autocracy) to 10 (full democracy). I will bin these 21 categories into two categories:

  • Full Democracy (polity score = 10)
  • Not a Full Democracy (polity score =9 to -10)

Thirty-two of the countries are full democracies and the remaining 123 are not.

Urbanization Rate (Possible Confounder)

The urbanization rate is measured by the share of the 2008 population living in an urban setting. This data was originally produced by the World Bank. The urban population is defined by national statistical offices. This variable was binned into those with a higher than average urbanization rate (55%+). 84 observations had a higher than average urbanization rate and 71 did not.

Results

The odds of having a higher than average income is almost 28 times higher for countries classified as full-democracies (OR=27.81, 95% CI=10.17 to 76.04, p=0.000).

However after adjusting for urbanization rate, the odds of having a higher than average income dropped to 17% for full democracies (OR=17.09, 95% CI=5.88 to 49.67, p=0.000). Urbanization was also significantly associated with high income, such that countries with a higher than average urbanization rate were significantly more likely to have higher than average per capita incomes (OR= 9.31, 95%CI=2.47 to 35.16, p=0.001).

Source Code

UPDATE: Jupyter notebook available on GitHub.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import statsmodels.api as sm
import statsmodels.formula.api as smf

# bug fix for display formats to avoid run time errors
pd.set_option('display.float_format', lambda x:'%.2f'%x)

df = pd.read_csv('gapminder.csv')

# convert to numeric format
df['incomeperperson'] = pd.to_numeric(df['incomeperperson'], errors='coerce')
df['polityscore'] = pd.to_numeric(df['polityscore'], errors='coerce')
df['urbanrate'] = pd.to_numeric(df['urbanrate'], errors='coerce')

# listwise deletion of missing values
subset = df[['incomeperperson', 'polityscore', 'urbanrate']].dropna()

# This function converts the polity score to a category
def convert_polityscore_to_category(polityscore):
    if polityscore == 10:
        return 1
    else:
        return 0

# Now we can use the function to create the new variable
subset['full_democracy'] = subset['polityscore'].apply(convert_polityscore_to_category)

counts = subset.groupby('full_democracy').size()
print(counts)

# Create a threshold
income_threshold = np.mean(subset['incomeperperson'])
print(income_threshold)

# Set binary flag that income per person is greater than the threshold
def income_higher_than_threshold(income):
    if income > income_threshold:
        return 1
    else:
        return 0

subset['high_income'] = subset['incomeperperson'].apply(income_higher_than_threshold)

counts = subset.groupby('high_income').size()
print(counts)

# Create a threshold
urbanization_threshold = np.mean(subset['urbanrate'])
print(urbanization_threshold)

# Set binary flag that urbanization rate is greater than the threshold
def urbanrate_higher_than_threshold(urbanrate):
    if urbanrate > urbanization_threshold:
        return 1
    else:
        return 0

subset['high_urbanrate'] = subset['urbanrate'].apply(urbanrate_higher_than_threshold)

counts = subset.groupby('high_urbanrate').size()
print(counts)

# logistic regression with society type
lreg1 = smf.logit(formula = 'high_income ~ full_democracy', data = subset).fit()
print (lreg1.summary())

# odd ratios with 95% confidence intervals
params = lreg1.params
conf = lreg1.conf_int()
conf['OR'] = params
conf.columns = ['Lower CI', 'Upper CI', 'OR']
print (np.exp(conf))

# logistic regression with society type and urbanization rate
lreg2 = smf.logit(formula = 'high_income ~ full_democracy + high_urbanrate', data = subset).fit()
print (lreg2.summary())

# odd ratios with 95% confidence intervals
params = lreg2.params
conf = lreg2.conf_int()
conf['OR'] = params
conf.columns = ['Lower CI', 'Upper CI', 'OR']
print (np.exp(conf))

Model Output

Logistic Regression Model 1 – Full Democracy

Optimization terminated successfully.
         Current function value: 0.389711
         Iterations 6
                           Logit Regression Results                           
==============================================================================
Dep. Variable:            high_income   No. Observations:                  155
Model:                          Logit   Df Residuals:                      153
Method:                           MLE   Df Model:                            1
Date:                Sat, 19 Dec 2015   Pseudo R-squ.:                  0.3091
Time:                        12:23:56   Log-Likelihood:                -60.405
converged:                       True   LL-Null:                       -87.436
                                        LLR p-value:                 1.944e-13
==================================================================================
                     coef    std err          z      P>|z|      [95.0% Conf. Int.]
----------------------------------------------------------------------------------
Intercept         -2.0523      0.284     -7.229      0.000        -2.609    -1.496
full_democracy     3.3253      0.513      6.478      0.000         2.319     4.331
==================================================================================
                Lower CI  Upper CI    OR
Intercept           0.07      0.22  0.13
full_democracy     10.17     76.04 27.81

Logistic Regression Model 2 – Full Democracy + Urbanization Rate

Optimization terminated successfully.
         Current function value: 0.342689
         Iterations 7
                           Logit Regression Results                           
==============================================================================
Dep. Variable:            high_income   No. Observations:                  155
Model:                          Logit   Df Residuals:                      152
Method:                           MLE   Df Model:                            2
Date:                Sat, 19 Dec 2015   Pseudo R-squ.:                  0.3925
Time:                        12:23:56   Log-Likelihood:                -53.117
converged:                       True   LL-Null:                       -87.436
                                        LLR p-value:                 1.246e-15
==================================================================================
                     coef    std err          z      P>|z|      [95.0% Conf. Int.]
----------------------------------------------------------------------------------
Intercept         -3.5056      0.636     -5.509      0.000        -4.753    -2.258
full_democracy     2.8388      0.544      5.216      0.000         1.772     3.905
high_urbanrate     2.2312      0.678      3.291      0.001         0.903     3.560
==================================================================================
                Lower CI  Upper CI    OR
Intercept           0.01      0.10  0.03
full_democracy      5.88     49.67 17.09
high_urbanrate      2.47     35.16  9.31

Democracy and Economic Well-Being

I have recently started the Data Management and Visualization course on Coursera.  As part of this course I will be conducting research and will be writing posts explaining my work.  This post is the first in the series of posts explaining my research project.

My Research Question & Hypothesis

For this research project I will be exploring what is the relationship between the level of openness of a society and the economic well-being of the citizens. My hypothesis is that countries with a more open society will have a higher level of economic well-being.

Data Used

The data I will be using is a subset of the Gapminder Project’s data. I measure the economic well-being by the 2010 per-capita GDP.  This data has been adjusted to replace purchasing power parity and is in constant 2000 US dollars.  I will measure the level of openness with the 2009 Democracy score (Polity).  This data comes from the Polity IV Project and is a summary measure of a country’s democratic and free nature. It ranges from -10 (an autocracy) to 10 (full democracy).

Literature Review

In Why Nations Fail1, Daron Acemoglu and James Robinson explain that economic prosperity depends above all on the inclusiveness of economic and political institutions.  This is in contrast to classical economics which explain the difference in economic well-being using factors like land, labor and capita.  I have previously read this book and hope that my research can confirm or deny their thesis.

I searched using Google Scholar for other literature on this topic.  John Gerring’s “Democracy and economic growth: A historical perspective”2 explained that recent studies appear to show that democracy has no robust association with economic growth (economic growth is viewed as the engine to increase economic well-being).  They argue that democracy must be understood as a stock, rather than a level, measure.

Next Steps

I plan on doing the analysis in Python and will be posting my work to GitHub.

Footnotes

1. Robinson, A. D., and R. Acemoglu. “Why nations fail.” The Origins of Power, Prosperity and Poverty, Nueva Y ork (2012).

2. Gerring, John, et al. “Democracy and economic growth: A historical perspective.” World Politics 57.03 (2005): 323-364.